How to Treat Non-Small Cell Lung Cancer with Actionable Mutations with Dr. Balazs Halmos

This is the Physical Activity Researcher Podcast, a podcast for researchers of sedentary behavior, physical activity, and sports. Join for a relaxed dialogue about research design, practicalities, and well, anything related to research. Learn from your fellow researchers useful and relevant information that does not fit into formal content and limited space of scientific publications. And here is your host, researcher and entrepreneur, Ollie Tickening. Yeah, so good point to consider allocating enough time. Having a good process where you kind of double check that the data is stored, data is maybe backed up before you erase it from the memory of the device. Do you have any experience just to share what can go wrong in this part? What kind of mistakes can you make in this stage when you are having 100 devices that you need to deal with? Let me think, mistakes might be. I can't remember, like I could be getting confused about. It could be even just momentary confusion, but which one you are already downloaded and which one you didn't. So you would have to put them in the docking station and check again to have this file. Do you have this data already? Do you have to download it again? If you are in hurry, as I said, all the devices look the same. If you might forgot which one you already downloaded, for example, then you have to check again and that's a waste of time. And then you want to double check again all the devices if you download everything. Even not downloading one might mean in some project losing one participant. If you think that you have many data points where you want to compare, if you lose two here and two here, and they are different ones, you start to go down quite quickly, so you need to be careful to do things correctly. And I think here again, like the newer devices with a cloud-based, you don't need to download the data. You don't need to initialize them. You can initialize them remotely. If you have a chance to use those, those might make it easier. Any other considerations for the practicalities with the measurement with the data? For the data handling or for the data analysis? I think we could go first with the data handling before going for the analysis. So again, make sure that you have a big memories to store the data and that these drives are ever back up. So usually university password protected drive are ever back up. So if anything happens, you'll be able to get the data. Do not store the data in an external drive as the only option. Because if that, if anything happens to that, your project is gone. I don't know if my experience, I heard about somebody that had some problem like this losing data. So make sure that you have a backup. You download data in a space and you have backup for all your data. That's the first thing. And basically, usually you have personal information that the data is linked to a person. So you need to be really careful with the storage. Exactly. So if an external drive wouldn't be ideal, normally you want some password protected company or university drive where you can only access it. And that should be covered in your ethical applications. And how would you do practically? Of course, ideal probably would be that you are in the premises of university and you download and it gets backed up in the cloud of the university and its password protected and so on. But if you are on the road, you don't have access to the university premises for any reason. Would you then kind of access the university computers via DPN or how would you do it practically? Again, this should be factored before. I mean, you don't want to have this kind of problems to solve in the way, especially if you could lose data or you don't know if the connection would be good in another place. So you need also to make sure that you can access the driver you store the data when you have to analyze the data in that time frame where you need to analyze and check the data. You need to be able to store it in the correct and secure place. So I think this again stresses the importance of piloting and I remember from my PhD measurements we used one brand which not very widely used and and basically it created like it measured heart rate or ECG curve, I think, an acceleration. And then it created when you got the results, it created like seven, eight different files and those will file formats that at least I didn't have even in the beginning of software that I can check that what is it they were data formats so. And then we didn't have time to look them well enough. Luckily we had the data but we hadn't really checked when we were doing measurements that what is it file is it actually valid information so you really want to check in the pilot thing that you go through and hopefully you even analyze the data that you you assert that you have the definitely that's key research trying to make sure that you know how to do. What you need to do yeah and and then maybe before we finish this one there's there's quite many settings in the in the devices that you need to or some devices have quite many settings others don't but you have epoch you have frequency you have measurement range. So what would you would you say say about this so that's always that always depends on what kind of assessment you decided to use but generally you need to set. Aspects like in accelerometers devices such as the epoch length the epoch would be that span of time where the accelerometer device averages the accelerometer output so for example in a five second epoch an accelerometer would do the average of the acceleration over five seconds. So again 60 minute epochs you would have the average over 60 seconds sorry 60 seconds epoch you would have the average over 60 seconds with children it is suggested to use epochs for physical activity assessment between one and 15 seconds because their physical activity they change physical activity quite. Addictively and really fast so they might have about of a vigorous physical activity for two seconds and then sit down and then jump and then so on while adults might have engage in physical activity in a bit more regularly when they start to do physical activity that would keep it going or well children physical activity you might want to consider. So physical activity assessments methods that have a smaller epochs because it was suggested that they are more accurate in assessing physical activity. And if I add to this like some newer devices that they are actually measuring raw data that you can actually have for example 10 11 20 20 hertz meaning that it's recording actually many samples per second so those are more accurate and enable going back kind of analyzing more accurately. But if your device needs epochs it's more accurate lower you put it but then the measurement time will get get sort right. So you mentioned about frequency you're talking about the frequency yeah right yeah so. Yes that's another key part you don't just choose the epoch but also the frequency within the epoch as you said so basically for example. they normally range between 10 and 100 hertz, so it means observation, so assessment per second. So normally the study is reporting methodologies to assess physical activity, such as cut points, for example, report specific epochs and specific frequency of data assessing, of data recording, so it's important to set the devices with the correct epoch assessment and correct frequency assessment. So for people who are not familiar with cut points, cut points are thresholds, which were developed to classify the levels, so the output for the accelerometers in different physical activity intensities. For example, if I get a really high acceleration, a really high acceleration, those could be categorized into vigorous physical activity. And so the cut points have been validated, should be validated in specific population, with specific epochs and frequency of assessment that should be used in a new research. And with count-based, this gets a bit more geeky, especially for count-based accelerators, cut points, using the wrong frequency could have bad effects, because when the accelerometer is assessing counts, having an higher frequency of assessment, which higher than the one that should be could lead to higher counts, because the accelerometer is assessing more acceleration every second, and could lead to wrong physical activity assessment, wrong physical activity output, because basically the accelerometer, the method was not tested to have that frequency, but yeah, this is a bit more geeky to say you need to check the frequency and the epoch. Yes, good points. And then I wanted to ask. Hello, excuse me! Could I ask one question before you? Yeah, yeah, sir, please, please go on. Do you work with children and physical activity? Okay, then you should listen to us. We wanted to tell you about this holiday bell animation that is so cool, and we understood from his funny explanations why humans and polar bears shouldn't sit all day. Fabian device, it is an accelerator, and I could tell you all the specs, but you just need to know it measures accurately, said until we behavior and physical activity, and is scientifically validated, but most importantly, using it feels like magic. Yeah, like magic. Ding dong dong! So, Lono@slashkids, please click it because it's important that kids all over the world can learn from polar bear. That's so awesome! Mom, can we have our polar bear as a pet? Please, Mom! So, you need to be careful to select the right ones and then based on what you have chosen also to analyze them, then correctly, that you have it. But I think this has been very nice, a lot of things about practicalities before going more into the analysis of the data. I think we will finish this part here. So, do you have anything to add into this discussion, maybe a summary in the end of practicalities? Practicalities of, do we know about the data analysis part? Or I will just say I think which is very important to remember when you work with physical activity data, that weather conditions and other factors like season, might influence physical activity and you need to account for these factors. Because when you assess data, so this comes later on. So, not in the data processing, but borrowing a data analysis, if you collect data, you might have very active children. If you collect the data about these children during rain days, their physical activity levels would be lower than normal. So, you want to account for that. And these are very, very important. Or even children tend to be more, in UK for example, children tend to be more active during summer and less active during winter. It is reported that during weekends, children tend to be less active than during the week days. This is again important. And these could guide the selection of the valid week criteria that I mentioned before, but I didn't explain valid day and valid week criteria. We might expand, do you want to expand on that? Yeah, please, please go on. So, it is also important and this is a huge effect on your final sample size and your and also on the quality of your data is selecting valid day and valid week criteria. So, generally, what do I mean with valid day for example? I mean that an observation of physical activity is considered representative of the actual daily physical activity of the child if the accelerometer collected at least a certain amount of hours per day. So, if the day, if the children is awake for, I don't know, 16 hours, you want to collect at least a certain amount of hours to make sure that your assessment is representative of this normal physical activity data. Of course, you want as much data as possible. If you could collect the whole day, all the hours that would be perfect, but in reality, that doesn't happen most of the time. So, it is suggested to have to collect at least 10 hours of awake time for each day and there's a good article, I think it's Miguel's 2018 explaining all this stuff. And you want to collect at least four days of valid. So, this is what is generally accepted to have at least four days of valid days to have a valid week of assessment of physical activity because you want enough of physical activity to be representative of the normal physical activity of the child. Some studies use less than four days, but that goes into risk of not being very representative of children's habitual physical activity. This is again, yeah, yeah, fundamental. Yeah, how would you consider then the valid day that, for example, if you have some children who have 10 hours of date per day and then others that have 16. So, the amount of physical activity can be quite different even if they would be actually in reality exactly same amount active. Would you report the results as a percentage of that measurement time? Like, for example, if you have one hour, it would be 10% of 10 hours, but much less from 16 hours. So, basically nowadays, most of the devices provide you with the possibility to to sort of infer the sort of calculate the physical activity for the missing hours. So, the missing wear time. So, if I have only 10 hours of valid wear time, so the children was wearing the device for only 10 hours. And the software can impute, can calculate what could have been the physical activity of the child over the hours where the child didn't wear the devices based on other days of measurement. So thanks to these methods, you can use, even if you have 10 valid hours, you can use the whole awake hours because thanks to the software, you impute the data so you have an idea about what could have been if you was wearing it, or she was wearing it all the time. But basically, for this, you need at least one long day that has that data. Yeah, but you need exactly, but you want to do this imputation of the missing data, only if you have enough data. So for example, 10 hours over the awake during the awake time, of course, the more is better and in our analysis of physical activity, we also accounted for valid wear time as a covariate. So you also want to account that for that because maybe differences in physical activity could be due to the amount of wear time rather than to other stuff. So it's important to account also for this factor. Yeah, very important points about the practicalities. And I think we covered quite a bit how to rotate the devices, how to prepare, how to allocate time, downloading, recharging, sanitizing, importance of piloting, and then about settings. Anything else you would like to add to this part? Yeah, maybe we didn't talk about wear time, great years, but there's generally algorithms that calculate wear time, valid wear time, which means there's accelerometer. So there's algorithms which calculate whether the participant was actually wearing the device or not. That's it, just clarification because we talked about. So basically based on it that if it's on the table, there's no accelerations experienced by the device and then if it's on a person even if laying on the soul for there's some movement, some accelerations. And based on that valid on that algorithm, you know you will have an estimated what was the actual wear time. And based on that, you will make your decision about keeping the data as a valid day and the day's as valid week and so on and doing your analysis. Yeah, I think this was a lot of useful information for our listeners who are planning to do measurements with children's product. There's quite many things to consider. And thank you, Matt, for taking the time for this recording. Thanks for inviting me. Hope it will be interesting for listeners and helpful. Yeah, and just as a reminder, Matt, they've just finished his PhD and is looking for a persistence. So if you have a project starting and maybe need someone who has the experience of running the measurements, doing the analysis and reporting and publishing the data, being in contact with Matt, they'll put the best way to contact you. Probably at the moment with my email, my personal email would be the best one should I give it to you maybe later or should I say my email? You can say it here. So, okay, it would be crottie.m or c-r-o-t-t-i dot m-c at gmail.com. Yeah, yeah. So email email to Matt if you have open persistence or maybe if you have a question. If you need suggestions. Yeah, if you have a question. If I can help you if I have time. Yeah. Yeah, he had been playing with hundreds of devices for a long time. So he knows a thing or a two. But anyway, thank you. I think this was a brilliant recording. I really enjoyed the chat. Thank you, Ali. Thanks for joining us this week on physical activity research report cards. If you like the show, make sure you never miss an episode by subscribing or following the show on Twitter. This podcast is made possible by listeners like you. Thank you for your support. If you found value in the show, we would really appreciate rating on Apple Podcasts or which ever app you use. Or if you would in real-old school way, simply tell a friend about the show. It would be a great help for us. We have a fantastic lineup of guests for forthcoming episodes, so be sure to tune in. Thank you all for your support and have a great day.